Research And Application Of Chinese Personal Relation Extraction Based On Deep Learning

As one of the important research directions in the information extraction task,relation extraction has been studied as a hot technology in the field of NLP in recent years due to its huge application value.Chinese personal relationship extraction is an important research direction of entity relationship extraction.The main purpose is to determine the character entities in the text,so as to judge the relationship between characters.It is of great significance to the construction of the personal relationship map.Among the existing research results related to Chinese character relationship extraction,the main problem reflected is the lack of public data sets,and to solve this problem,people mainly use the method of remote supervision to establish the annotation corpus required for training.The advantage of this method is that it solves the problem of lack of training data.At the same time,this method also has shortcomings.When using the remote supervision method to construct a data set,a large amount of noise data will be introduced,which will affect the training of the model and lead to poor relationship extraction results.In response to this problem,this paper proposes an effective optimization method through the analysis of existing methods,and conducts an in-depth exploration of the application of deep learning in the field of character relationship extraction.The main research contents of the paper are as follows.(1)In view of the lack of data sets,this paper obtains character relationship pairs from the Chinese knowledge bases "Baidu Baike" and "CN-DBpedia",then crawled some Baidu Baike character profile data,combined with some network public data sets,and applied remote supervision This part of the data is marked by manual marking method,and a Chinese character relationship extraction data set(11280 text data)is constructed.(2)In this paper,four different deep learning models,Bi LSTM,Att-Bi LSTM,Bi GRU and BERT+Bi GRU+Att,are used to implement Chinese character relationship extraction.By analyzing the results of Bi LSTM model and Att-Bi LSTM model,it is verified that the attention mechanism can effectively improve the effect of relation extraction.Compared with the Att-Bi LSTM model using Word2 vec word embedding,the precision rate,recall rate,and F1 value of the extraction results of the BERT+Bi GRU+Att model are improved.Therefore,it can be concluded that using the method of training word vectors by BERT can obtain word vectors with richer semantic information,which is of great help to the extraction results of the model.(3)Based on the neural network model constructed above,in view of the current market demand for open-source character relationship extraction system,combined with Vue,Spring Boot and Flask and other mainstream frameworks,this paper designs and builds a Chinese character relationship extraction system to realize the whole process service system from relationship extraction,relationship storage and relationship visualization,which provides convenience for users.The system mainly includes the functions of relationship extraction,data management,relationship visualization and so on.